Kafka Consumer and Producer Interceptors for tracing and report to Zipkin.
These interceptors could be plugged into Kafka applications via classpath configuration.
The purpose of this instrumentation is to enable tracing for Kafka Connect and other off-the-shelf components like Kafka REST Proxy, ksqlDB, etc. For more complete tracing support, check Brave instrumentation for Kafka Clients and Kafka Streams.
Producer Interceptor create spans when sending records. This span will only represent the time it took to
execute the on_send
method provided by the API, not how long to send the actual record, or any other latency.
Add Interceptor to Producer Configuration:
producerConfig.put(ProducerConfig.INTERCEPTOR_CLASSES_CONFIG, Collections.singletonList(TracingProducerInterceptor.class));
//or
producerConfig.put("interceptor.classes", "brave.kafka.interceptor.TracingProducerInterceptor");
Consumer Interceptor create spans on consumption of records. This span will only represent the time it took execute
the on_consume
method provided by the API, not how long it took to commit, or any other latency.
consumerConfig.put(ConsumerConfig.INTERCEPTOR_CLASSES_CONFIG, Collections.singletonList(TracingConsumerInterceptor.class));
//or
consumerConfig.put("interceptor.classes", "brave.kafka.interceptor.TracingConsumerInterceptor");
Key | Value |
---|---|
zipkin.sender.type |
Sender type: NONE (default), KAFKA , HTTP |
zipkin.encoding |
Zipkin encoding: JSON (default), PROTO3 . |
zipkin.http.endpoint |
Zipkin HTTP Endpoint sender. |
zipkin.kafka.bootstrap.servers |
Bootstrap Servers list to send Spans. if not present, bootstrap.servers (Kafka Client property) is used. |
zipkin.local.service.name |
Application Service name used to tag span. Default: kafka-client. |
zipkin.trace.id.128bit.enabled |
Trace ID 128 bit enabled, default: true |
zipkin.sampler.rate |
Rate to sample spans. Default: 1.0 |
zipkin.kafka.* |
Use this prefix to override any kafka producer property |
Required software available:
- Docker and Docker Compose available.
- JDK 11+
Start by building libraries and run Docker Compose:
make docker-up # will build jars and start containers
# or
./mvnw clean package
docker-compose up -d
Create database tables:
make pg-table
Once table is created deploy source and sink connectors:
make kafka-connectors
Create a KSQL table and select to wait for results:
make ksql-stream
make ksql-select
Insert new rows:
make pg-row
Go to http://localhost:9411 Check the traces.
Traces should look like this:
Search:
Trace view:
Dependencies:
All artifacts publish to the group ID "io.zipkin.contrib.brave-kafka-interceptor". We use a common release version for all components.
Releases are at Sonatype and Maven Central
Snapshots are uploaded to Sonatype after commits to master.